Meir Statman - The Retirement Advice Centre

Transcription

Meir Statman - The Retirement Advice Centre
Dollar-cost
averaging may not be rational behavior, but it is perfectly normal behavio1:
Meir Statman
I
nvestors with cash that is destined for stocks often
use a dollar-cost averaging plan. They divide the
cashinto segments, and convert one segment at a
time from cashto stocks according to a predetermined schedule. The alternative to dollar-cost averaging is lump-sum investment.
The popularity of dollar-cost averaging can be
traced back at leastto the 1940s. (See, for example, discussions in Ketchum [1947], Solomon [1948], and
Weston [1949].) And that popularity has never waned.
For example, Clements [1994] writes in a Wall
StreetJournal column "aimed at ordinary investors who
want to get their finances going in the right direction":
Tumbling stock and bond prices can seem a lot
less painful if you plan to buy more. One of the
best ways of doing that is dollar-cost-averaging,
which involves shoveling, say, $100 into the
market every month, no matter what is happening to stock and bond prices (p. C1).
MEIR
STATMAN
is professorof
finance at the Leavey School of
Businessof SantaClara University in
SantaClara (CA 95053).
70
A BEHAVIORAL FRAMEWORK FOR. DOLLAR.-COST AVERAGING
While popular, the practice of dollar-cost averaging is inconsistent with standard finance. This has
been demonstrated by Constantinides [1979], who
shows, within a theoretical framework, that dollar-cost
averagingplans are suboptimal. It has also been demonstrated by Rozeff [1994], who shows using simulation
that dollar-cost averaging is suboptimal.
An analysisof dollar-cost averaging is important
for at leasttwo reasons,one related to an understanding
FALL
1995
of the behavior of investors, and the other related to the
effects of investor trading on security prices. Standard
finance is a positive theory, a theory that makespredictions about the financial behavior of individuals and
about the outcomes of the interactions betWeenindividuals in financial markets. The practice of dollar-cost
averaging is prominent, and the inconsistency betWeen
the practice of dollar-cost averagingand the predictions
of standard finance is too glaring to be ignored.
Moreover, an understanding of the persistence of dollar-cost averaging provides insights into broader questions, such as the overall construction of portfolios.
This article offers a behavioral framework that is
consistent with the persistence of dollar-cost averaging.
I describe the roles of four behavioral elements in the
attraction of such plans: prospect theory, aversion to
regret, cognitive errors, and self-control (behavioral life
a relationshipthat hol~ without exception: at
anypoint aftera flucroationin securitypricesthe
averagecostof total sharesheld is lessthan the
averageprice of the shares(pp. 251-252).
Weston exposesthe irrelevance of this fact: "The
crucial testis whether the sharesheld can at any time be
sold at a gain. For this to be possible,averagecost must
be lessthan the current market price per share" (p. 252).
Similarly, Sharpe [1981] notes that while it is
mathematically interesting that the average price per
share paid by a dollar-cost averaging investor is lower
than the averageprice per share, it has no economic
significance. Sharpe shows that while high volatility in
stock prices corresponds to large differences betWeen
the averageprice per sharepaid by a dollar-cost averaging investor and the averageprice per share, dollar-cost
averaging does not change uncertainty from vice to
virtue. The passageof time since Weston's 1949 article
and Sharpe's1981 book seems to have done litde to
dampen enthusiasmfor dollar-cost averaging
The world of standard finance is the world of
frame invariance. Investorscare about cashflo\vs, but are
indifferent among frames of cashflows. The pricing of
options is a good example. The price of a call option on
a stock is determined by the fact that the cash flows of
the option can be replicated by the cashflows of a particular dynamic combination of a bond and the underlying stock. The fact that in the first casecashflows are
described in terms of options, while in the second cash
flows are described in terms of bonds and stocks is irrelevant to investorsin a world of frame invariance.
Although the literature of standard finance has
no relevant role for framing, the behavioral literature is
replete with studies on the effects of frames on choice.
The effect of frames is central in prospect theory, a positive theory of choice by Kahneman and Tversky
[1979], and with it I begin the construction of the
behavioral framework within which dollar-cost averaging takes place.
cycle theory).
This work is part of a stream of work that
describes the behavior of investors and the outcomes of
their interaction in financial markets. Earlier work
describes preferences for dividends (Shefrin and
Statrnan [1984]), the reluctance to realize losses[1985],
the susceptibility to cognitive errors and the preference
for stocks of "quality" companies [1986, 1995b], the
design of securities [1993], the pricing of securities
[1994], and the construction of portfolios [1995a].
"Behavioral investors" make choices in a systematic, if suboptimal, fashion. This is not to advocate
the selection of suboptimal portfolios. But a positive
theory must be consistent with the behavior of many, if
not most, individuals.
Some standard finance investors (and academics)
think that behavioral investors can be easily educated
to overcome their limitations. But even if they are
right in their prescription, standard investors will be
ineffective as teachers if they rnisperceive their students. Behavioral investors are numerous, and they are
difficult to educate. The difficulty in the task of education is illustrated in Weston's [1949] and Sharpe's
[1981] efforts.
Dollar-cost averaging calls for investing the PROSPECT THEORY
same dollar amount, rather than the same number of
Choices of standardfinance investors conform tc
shares, each period. Thus, a dollar-cost averaging
investor buys more shares when the price is low than expected utility theory. Choices of "behaviora' !
investors" conform better to prospect theory: Prospect
when the price is high. As Weston [1949] writes:~
theory investors evaluate their choices in terms of the
potential gains and losses relative to reference points,
In the usual exposition of the principle of dollarwhile standardinvestors evaluate their choices in terms
cost-averaging, its merit is urged on the basis of
)
FALL 1995
THEjOURNALOF PORTFOUOMANAGEMENT71
EXHIBIT lA
STANDARD UTll.ITY
Kahneman and Tversky find that 84% of subjects chose A1, the sureamount, in the first problem set.
Yet, 69% of subjectschose B2, the gamble, in the second problem set. This pattern of choice is puzzling
within standard finance, because standard finance
investors base their decisions on net cashflows and are
never confused by frames. Yet, problem sets1 and 2 are,
in fact, identical in net cashflows.
Observe that once the initial $1,000 is integrated into the choice between Al and Bl in problem 1, the
overall choice is between:
FUN C TI 0 N
Utility
Total Wealth
A3: A sure gain of $1 ,500 (the sum of the initial
$1,000 and the sure $500), and
B3: A 50% chance to gain $2,000 and a 50%
chance to gain $1,000.
Similarly, once the initial $2,000 is integrated
into the choice between A2 and B2 in problem 2, the
overall choice is between:
A4: A sure gain of$1,500, and
of net cash flows (total wealth). Moreover, while stanB4: A 50% chance to gain $2,000 and a 50%
dard investors are always risk-averse, prospect theory
chance to gain $1,000.
investors have an S-shaped value function over gains
and losses that displays concavity (risk aversion) in the The tWo problems are identical in net cashflows.
domain of gains and convexity (risk-seeking) in the
domain of losses. (See Exhibits lA and lB.)
EXHIBIT IB
The origins of prospect theory are in Markowitz
PROSPECT FUNCTION
[1952], but its development is the work of Kahneman
and Tversky [1979]. To understand the features of
prospect theory, consider an experiment by Kahneman
and Tversky. One group of subjectsreceivesproblem 1:
1. fn addition to whatever you own, you have
been given $1,000. You are now askedto choose
betWeen:
A1: A sure gain 0£$500, and
B1: A 50% chance to gain $1,000 and a 50%
chance to gain nothing.
Another group of subjects receives problem 2:
2. In addition to whatever you own, you have
been given $2,000. You are now askedto choose
between:
~: A sure loss of $500, and
B2: A 50% chance to lose $1,000 and a 50%
chance to lose nothing.
72
A BEHAVIORAL FRAMEWORK FOR. DOLLAoR-COST AVERAGING
FALL 1995
ioral way, the problem shows a gain in all casesexcept
when the stock price never changes. It is absolutely
true that the behavioral frame is misleading. It is equalPrice per
Amount
Number of
ly true that the behavioral frame persists.
Period
Share
Invested
SharesBought
Unforrunately. there is no comprehensive theory
that
explains what makes some frames more com20
$50.00
51,000
pelling
the
others. (See Fischhoff [1983].) However, the
2
80
$12.50
51,000
persistence of the behavioral frame of dollar-cost aver100
Total
$2,000
aging is hardly unique. Consider the public discussion
about derivatives. Some finance practitioners and aca$2,000/100
= $20
Average CoSt of SharesHeld:
demics frame derivatives in the standard finance way
and know that derivatives can be used with equal effecAveragePrice per ShareOver
(50 + 12.5)/2 = $31.25 tiveness to increase risk or to reduce it. But framing
the Two Periods:
derivatives such that they always increase risk is a common practice, hardly limited to politicians.
A prominent fearure of dollar-cost averaging is
Most of Kahneman and Tversky's subjects could
not possibly be standard finance investors. Rather, they that it is recommended with equal force to investors
are behavioral finance investors. Prospect theory postu- with cashwho consider converting cash into stock and
lates that tWo distinct cognitive operations lead to investors \vith stock who consider converting stock into
choice, and that these tWo operations are sequential. cash. This fearure is useful in highlighting the difference
First is framing into mental accounts. Second is the in framing and choice berween standard finance and
behavioral finance.
application of specific decision rules to the accounts.
Constantinides [1979], who analyzes dollarThe initial amount, $1,000 in problem 1, is
stripped away and framed into a separate account. cost averaging within the framework of standard
Problem 1 is then framed in terms of gains and losses finance, writes:
relative to a reference point of zero. The concave porWhere, then, does the intUitive rationale of doltion of the prospect function in the domain of gains
lar-cost-averaging fail? Its rationale is that the
leads to a preference of the sure $500 gain over the
investor replaces one major gamble on a tempogamble, a choice consistent with risk aversion. In probrary shift of prices by a nwnber of smaller gamlem 2, the convex portion of the prospect function in
bles and thus diversifies risk. The fault of this
the domain of lossesleads to a preference of the gamargument is misrepresentation of the state of the
ble over the sure $500 loss, a choice consistent with
E.."XHIBIT 2
Dollar-Cost Aver:lging
world, before a decision is made. Dollar-costrisk-seeking.
averaging implies that an investor with all his
Consider now framing and choice in the context
endoWment in assetA is in some way different
of dollar-cost averaging. Imagine an investo~ who
from an investor with all his endowment in asset
divides $2,000 in cash into tWo segments of $1,000
B, but otherwise identical. Dollar-cost-averaging
each, investing one in period 1 and the second in periignores
the simple fact that the latter investor
od 2. The price per share of stock in period 1 is $50,
may cosdesslyconvert his endowment from asset
and it turns out that the price per share in period 2 is
A to asset B before he considers the optimal
$12.50. The data are presented in Exhibit 2.
investment decision. Both investors face the
Framing the problem in the standard finance
same
prospects irrespective of the composition
way, the investor started with $2,000, and now has 100
of their endowment, and any claims of gambles
sharesworth $12.50 apiece for a total of $1,250. The
on temporarily overpriced or underpriced prices
investor has a clear loss.
are simply fallacious (pp. 447-448).
Framing the problem as the proponents of dollar-cost averaging would have it, the investor bought
Imagine two investors, A and B, who are identhe shares at an average cost of $20, while the average
tical
except
that A has $1,000 in cashand B has $1,000
price per share over the tWo periods was $31.25. The
investor has a clear gain. Indeed, framed in the behav- in stocks. A faces a choice between keeping his wealth
FALL 1995
THEJOURNALOF POI\.TFOUO MhNhGEMENT 73
in cash or converting it into stock while B faces a
choice betWeen keeping her wealth in stock or converting it into cash. Framed in the standard finance
way, the choice problems of A and B are identical
because B can cosdessly convert her initial stock
endowment into cash. Therefore, their choices are
predicted to be identical.
The frames and choices of A and B are likely to
be different within the framework of behavioral
finance. The tWo are identical in their beliefs, so they
agree that the return on cashis zero, and that the value
of stocks at the end of the period will, with equal probabilities, either increase to $1,300 or decreaseto $860.
The expected gain on stocks is $80, while the expected gain on cashis zero.
How would A frame the choice? Assume that
the reference point for A is the $1,000 in cash, a position he has adapted to, and that he frames the choice in
terms of gains and lossesrelative to the $1,000 reference
point. If so, the choice is betWeen:
Cash A. A sure gain of zero, and
Stock B. A 50% chance to gain $300 and a
50% chance to lose $140.
Assume that the reference point for B is $1,000
in stocks, a position she hasadapted to. If so, the choice
is betWeen
Cash
A
Stock B.
A 50% chance for an (opportunity)
gain of $140 and a 50% chance for
an (opportunity) loss of$300, and
A sure (opportunity) gain of zero.
The problems faced by A and B are framed
differently, and the choices are thus likely to differ.
The concavity of the prospect function in the
domain of gains, and the convexity of the prospect
function in the domain of losses, is likely to cause A
to hold onto his cash, and it is likely to cause B to
hold onto her stock.l
The purported advantages of dollar-cost averaging involve, as Constantinides demonstrates, misleading frames. Framed in the standard finance way,
a dollar-cost averaging investor only replaces one
major gamble, embedded in a lump-sum investment,
with a number of smaller gambles, embedded in dollar-cost averaging. But frames are important, and
they affect choice.
74
A BEHAVlOIV.L FRAMEWORK FOR DOLL.o.R-COST AVERAGING
AVERSION TO REGRET
The purchase of stock for $1,000 will result in
$1,300 at the end of the period, or it will result in $860.
The monetary gain is $300, and 'the monetary loss is
$140, but monetary gains and losses are not all that
affects choice. The joy of pride and the pain of regret
matter. Kahneman and Tversky [1982] describe regret
as the frustration that comes, ex post, when a choice
results in a bad outcome.
If the $1,000 purchase of stocks results in
$1,300, the $300 monetary gain is supplemented with
the pride that comes from what is framed as buying
$1,300 worth of stock for $1,000. If the $1,000 purchase of stocks results in $860, the $140 monetary loss
is supplemented with the regret that comes from what
is framed as buying $860 worth of stock for $1,000.
The distinction between 1) gains and losses in
terms of money and 2) gains and losses in terms of
pride and regret is akin to Thaler's [1985] distinction
between acquisition utility and transaction utility. In
Thaler's framework, the total utility of the purchase is
composed of acquisition and transaction utilities.
Acquisition utility depends on the difference between
the value of the product and the outlay. Transaction
utility depends on the "bargain" value of the purchase.
In this framework, the bargain value corresponds to
pride and regret.
Standardfinance investors are affected by neither
pride nor regret. Pride and regret, however, do matter
to behavioral investors. If the joy of pride is equal to the
pain of regret, behavioral investors who choose stock
over cash without considerations of pride and regret
would not alter their choice once pride and regret are
introduced. If the pain of regret is sufficiently larger than
the joy of pride, however, behavioral investors would
choose to keep their holdings in cash rather than suffer
the pain of regret that will come if stock prices decline.
Kahneman and Tversky note that there is a close
associationbetween regret and the level of responsibility for a choice. Actions taken under duress entail little
responsibility and bring little regret. Following a rule is
one way to reduce responsibility. Choice under a strict
rule is choice under duress. Dollar-cost averaging
involves a strict rule that specifies amounts to be invested at particular points of time. The ability of a dollarcost averaging plan to reduce responsibility is especially
helpful for investors who are concerned about their
exposure to regret?
FALL 1995
COGNITIVE ERRORS AND SELF-CONTROL
Dollar-cost averaging is a non-sequential or
non-contingent investment policy. The non-sequential
nature of dollar-cost averaging is manifested in a commitment at the initiation of the plan to invest a particular amount in each subsequent period, regardlessof
any information that might become available after the
initiation of the investment plan. Constantinides
[1979] notes that the non-sequential nature of dollarcost averaging is considered by its proponents as the
key to its success.
Constantinides [1979] shows that dollar-cost
averaging is dominated by a sequential optimal investment policy, a policy that takes into account information that arrives after the initiation of the investment
plan. He adds that, in light of this result, it seemsironic that proponents of dollar-cost averaging go to great
lengths to emphasize that investors must have the
courage to ignore new information as they follow the
inferior non-sequential investment policy.
A policy that is suboptimal within standard
finance might nevertheless be attractive to behavioral
investors. One advantage of the non-sequential nature
of dollar-cost averaging for behavioral investors is that
the non-sequential rules of dollar-cost averagingreduce
responsibility and regret. But the advantage of following rules extends beyond a reduction in responsibility.
The rules of dollar-cost averagingserve to combat lapses in self-control as cognitive errors influence investors
to terminate their investment plans.
To understand the roles of self-control and cognitive errors, consider the description of dollar-cost
averaging by Cohen, Zinbarg, and Zeikel [1977], quoted by Constantinides [1979]:
The
important
ule-to
buy,
thing
even
is
though
to
stick
the
price
To
which, engagepsychologically,
in
dollar-cost-averaging
is usually
you
must
continue
have
both
buying
the
in
a
funds
declining
and
to
your
sched-
keeps
falling,
hardsuccessfully,
to
do
the
courage
market
to
The tendency of investors to eXtrapolate recent
trends in stock prices is a reflection of representativeness,a cognitive error, and that tendency is well-documented. For example, Solt and Statman [1988] find that
investment advisors become optimistic about the
prospe~tsof stocks after increases in stock prices and
pessimistic after declines. They also find that there is no
relationship betWeenthe sentiment of investment advisors at one particular time and the performance of the
stock market in the subsequentperiod.
Suppose a dollar-cost averaging investor startS
the investment plan with the expectation that there is
an equal chance for an up-market or down-market in
the coming period. Once severaldown-periods occur,
the investor revisesthe probabilities so that the probability of a down-market is higher. The investment plan
that was attractive by the old probabilities might no
longer be attractive by the new ones, and the investor
might choose to abandon the plan and stop buying
stocks. Here is where the self-control role of dollar-cost
averagingis most important.
Investors who allocate funds between savings
and consumption often face difficulties because consumption is tempting. Rules are useful in enforcing a
savings plan. Shefrin and Statman [1984] show how
rules such as "consume from dividends, but don't dip
into capital" help investors manage the self-control
problem when a myopic "agent" within the individual wants to consume now, but a forward-looking
"principal" considers savings for the future as well as
current consumption. "Don't dip into capital" is a
rule that the principal uses to constrain the consumption of the agent.
The task of the principal in enforcing savingsis
especially difficult after a period of losses, which is
when the strict rules of dollar-cost averaging are most
effective. The rules of dollar-cost averaging help
investors "continue buying in a declining market when
prospectsmay seembleak."
CONCLUSIONS
when
Investorswho employ dollar-cost averaging have
their wealth in one asset, such as cash, and consider
As Cohen. Zinbarg, and Zeikel note, investors transferring it into another asset,such as stock. They
find it difficult to continue to buy stocks following can transfer wealth from one asset to the other in a
stock price declines. But why do investors find it diffi- lump sum. Instead, they transfer wealth in increments
cult? The answer is that investors generally believe that over time according to a predetermined plan.
It has been known at least since Weston [1949]
recent trends in stock prices will continue.
prospects
FALL
1995
may
seem
bleak.
THEjOURNAL OFPORTFOLIOMANAGEMENT75
that the practice of dollar-cost averaging is inconsistent
with standard finance. Yet dollar-cost averaging seems
as popular today as ever. The persistence of dollar-cost
averaging is an embarrassment to the role of standard
finance as a positive theory of financial behavior.
Dollar-cost averaging is significant even if it is
followed only by small investors, as the aggregate of
small investors is large. Moreover, the early literature on
dollar-cost averaging, such as Cottle and Whitman
[1950], suggeststhat dollar-cost averaging plans were
then popular among institutional investors. There is no
evidence of a decline in that popularity.
Dollar-cost averaging is consistent with the positive framework of behavioral finance. I have described
the role of four elements of the theory: prospect theory, aversion to regret, cognitive errors, and self-control.
Choices that involve transfers of wealth among assets
are framed and evaluated within prospect theory to
show that dollar-cost averaging transfers are appealing
to investors who find lump-sum transfersunappealing.
Considerations of pride and regret affect transfers of
wealth among assets.The susceptibility to cognitive
errors, in particular the tendency to extrapolate recent
trends in stock prices, explains why investors find it difficult to continue dollar-cost averaging plans after a
period of stock price declines, and the need for rules to
facilitate self-control explains the non-sequential nature
of the rules that govern dollar-cost averaging.
Dollar-cost averaging joins financial products
such ascovered calls and LYONs, described by Shefrin
and Statman [1993], as products that fit poorly with
standard finance yet fit well with behavioral finance.
Indeed, it belongs in the general area of portfolio construction. Shefrin and Statman [19954] show that
investors generally construct portfolios in ways that
deviate from standard finance theory but are consistent
with behavioral finance.
Much of dollar-cost averaging takes place in a
framework where choice is not explicit. A feature of
the implementation of defined-contribution pension
plans, such as401 (k)s, is that employers and employees
contribute cash to the pension plan on each payroll
date, and the cash contribution is converted on that
date into stocks or bonds. Any choice between lumpsum and dollar-cost averaging in defined-benefits pension plans, however, is only implicit becauseemployees
are not given an explicit choice between contributions
in portions over the cause of a year and a lump-sum
contribution at a point during the year.
76
A BEHAVIORAL FRAMEWORK FOR DOLLAR-COST AVERAGING
Suppose an employer does offer employees a
choice between dollar-cost averaging contributions
coinciding with payroll days during a year and a onetime lump-sum contribution of the total annual
amount so that the present values of the cash flows in
the two options are identical. The prediction of standard finance is that employees will be indifferent
between the dollar-cost averaging option and the
lump-sum option. The prediction of behavioral finance
is that employeeswould prefer the dollar-cost averaging
option. While I know of no employer who offers such
a choice at present, it should be possible to test this
hypothesis in an experiment.
While my focus here is on investor behavior,
not on security prices, the practice of dollar-cost averaging has important implications for pricing. It is by
now well established that investment flows, even in the
absenceof information, affect prices. For example, the
work of Warther [1994] reveals a strong link between
cashflows into and out of mutual funds and the returns
to stocksheld by the funds. Investors who practice dollar-cost averaging are more likely than other investors
to continue to buy stocks after a period of declines in
stock prices and less likely to accelerate buying after a
period of increases in stock prices. I hypothesize that
an increase in dollar-cost averaging leads to a decrease
in volatility.
Dollar-cost averaging is indeed suboptimal
within the choice set facing a fully rational investor in
standard finance. But the interpretation of rationality
is a delicate task. Consider, for example, the equity
risk premium (Mehra and Prescott [1985]). The existence of an equity premium puzzle suggests that
investors invest too litde in stock, and investment
advisors might wish to guide their clients to convert
some cashinto stock.
Compare an advisor who counsels a client to
convert cash into stock in a lump sum to an advisor
who counsels the client to use dollar-cost averaging.
Lump-sum conversion from cash to stock might be
optimal, but such conversion is unappealing to investors
who are deterred from action as they contemplate the
regret that they will e>..-perienceif the stock market
were to crashassoon asthe cashis converted into stock.
Dollar-cost averaging is indeed a second-best solution,
but it might start an investor on a road that leads to allocation of a portion of wealth to stocks.
As Samuelson [1994] writes about dollar-cost
averaging, he notes that it is one of dozens of rules
FAll
1995
whose merit has nothing to do with improving risk- Fischhofl", Baruch. "Predicting Frames." Journal of Experimental
adjusted returns or mean-variance optimization. He Psychology,Learning,,Wemoryand Cognition, 9, 1 (January 1983), pp.
103-116.
adds that, at least for fiduciary trustees, using such rules
is a blunder, if not a crime. Yet the fact that many com- Graham,).R.. and C.R. Harvey. "Market Timing Ability and Volatility
mon investment rules are inconsistent with standard Implied in Investment Newsletters' Asset Allocation Recommendations." Working paper, University of Utah. 1995.
finance is evidence that standard finance does not do
\vell asa positive theory.
K.1hneman,D., and A. Tversky. "Prospect Theory: An Analysis of
1979,pp. 263-291.
Standard finance is inconsistent with the exis- Decision Making Under Risk." Econometrica,
tence of an investment advising industry where rules -."The
Psychology of Preferences." ScientificAmerican,246 (1982)
such as dollar-cost averaging are a mainstay. Standard pp.167-173.
finance is inconsistent with the existence of a mutual
fund industry where, on average,money managers fail Ketchum, M.D. "Investtnent Management Through Fornluh Timing
Plans.Journal ofBusiness,
XX Guly 1982),pp. 157-158.
to outperform indexes (see Malkiel [1995]), and standard finance is inconsistent with the existence of an Malkiel, B. "Returns From Investing in Equity MutUal Funds 19711991."journal of Finance,50 Gune 1995),pp. 549-572.
investment newsletter industry where newsletter \vriters provide uselessassetallocation advice (see Graham l\-1.1rkowitz,H. "The Utility of Wealth." Journal ofPoliticalEconomy.60
and Harvey [1995]).
(1952), pp. 151-158.
It might be time to move on to a positive theMehra, R., and E.C. Prescott. "The Equity Premium Puzzle." journal
ory that is consistent with the evidence, and to ~f ,\1onetaryEr.onomics,
40, 2 (1985), pp. 145-161.
remember that a normative theory is useless if
investors cannot be persuaded to follow it. Pre, Gordon. "Minimax Policies for Selling an Asset and Dollar
Science,17,7 (March 1985), pp. 379-393.
Meanwhile, I offer an hypothesis. The practice of dol- Averaging." ,Wanagement
lar-cost averaging will persist.
Roll, R. "A Mean/Variance Analysis of Tracking Error. " journal of
PortfolioMdnagement,
Summer 1992, pp. 13-22.
ENDNOTES
The author thanks Hersh Shettin and Atulya Sarin for helpful discussions,and the Dean Witter Foundation for financial support.
tThe choice problem ofB is equivalent to the choice problem of a money managerwhose performance is evaluated relative to a
benchmark identical to B's portfolio. The benchmark servesasthe reference point. and gainsand lossesare measuredrelative to it. (SeeRoll
[1992] and Clarke, Krase. and Statman [1994].)
2For a "mini-max" policy. where considerationsof regret
lead investors to choose dollar-cost averaging,seePye [1972].
Rozeff. Michael. "Lump-Sum Investing Versus Dollar-Averaging."
Journal ofPortfolioManagement,
Winter 1994, pp. 45-50.
Samuelson, Paul. "The Long-Te= Case for Equities." journal
PortfolioJWanagement.
Fall 1994, pp. 15-24.
of
Sharpe,W.F. Investments.
Englewood Cliffs, NJ: Prentice-Hall, 1981.
Shemn, Hersh, and Meir Statman. "Behavioral Aspectsof the Design
and Marketing of Finmcial Products." Financial Management,22, 2
(1993),pp. 123-134.
REFERENCES
-."Behavioral
Capital AssetPricing Theory." Journal ofFinancialand
Quantitatill~Analysis,29, 3 (1994). pp. 323-349.
Clarke, Roger G., Scott Kr:ISe,and Meir SUtman. "Tracking Errors,
Regret, and Tactical AssetAllocation." Journalof Portfolio,Wanagement, -."Behavioral
Portfolio
Spring 1994, pp. 16-24.
University, 1995a.
Clements,Jonathan. "Expect Some Bumps, and Hang On For me Long
Haul." Wall StreetJournal, October 24, 1994, p. C1.
Cohen, Jerome B., Edward D. Zinbarg, and Arthur Zeikel. Investment
Analysisand PortfolioJWanagement.
Homewood, IL: Irwin. 19ii.
Constantinides, G.M. "A Note on the Suboptimalicy of Dollar-Cost
Averaging asan Invesanent Policy." Journal ofFinancialand Quantitarive
Analysis,XIV, 2 (June 1979),pp. 443-450.
Cotde, C. Sidney, and W.T. Whianan. "Focmula Plans and the
InstitUtional Investor." Harvard BusinessReview,28, 4 auly 1950),pp.
84-96.
FALL
1995
Theory.
"
Working
paper,
Santa
Oara
-."The
Disposition to Sell Winners Too Early and Ride Losers
Too Long: Theory and Evidence." journal ofFinanle, 40 (July 1985),pp.
777-790.
-."Explaining
Investor Preference for Cash Dividends."]oumal
FinancialEconomics,
13 (June 1984), pp. 253-282.
of
-."How
Not to Make Money in the Stock Market." Psychology
Today,20, 2 (February 1986),pp. 52-57.
-,
"Making Senseof Beta, Size, and Book-to-Market," journal of
PortfolioJV1anagement,
Winter 1995b, pp. 26-34.
THEJOURNALOF PORTFOUO MANAGEME.'IT77
Solomon, Ezra. ..Are Fon1lula Plans What They SeemTo Be?" Journal
ofBusiness,XXI (April 1948), esp. pp. 96-97.
Solt, Michael E., and Meir SUtman. "How Useful is the Sentiment
Index?" FinancialAnalystsJoumal,September/October 1988,pp. 45-55.
Thaler, Richard.
"Mental Accounting and Consumer Choice.
MarketingScience,
4, 3 (Summer 1994),pp. 199-214.
Wanher, V.A. "Mutual Fund Flows and Security Returns." Working
paper, University of Southern California, 1994.
Weston. ].F. "Some Theoretic4l Aspects of Formula Timing Plans."
jo/4mal of BusinESs,
XXII, 4 (October 1949), pp. 249-270.
In addition to our Editorial Advisory Board, the following
have been most helpful in providing reviews of manuscripts
submitted to us in recent months:
Clifford Asness (Goldman SachsAsset Management)
Josef Lakonishok (Unive~ity of Illinois)
Michael Rosenberg (Merrill Lynch)
78
A BEHAVIORAL FRAMEWORK FOR DOU.AR-COST AVERAGING
FALL
1995